Scholarlink

<p>Stone column is an effective and sustainable technique of ground improvement, but the prediction of the bearing capacity of foundation built on stone column reinforced soils is still difficult to achieve because of uncertainties in the conventional method. In this study, the performance of four Machine Learning models namely Radial Basis Neural Network (RBNN), Generalized Regression Neural Network (GRNN), Random Forest (RF) and Support Vector Regression (SVR) is compared to develop a prediction model for the bearing capacity (qu) of floating stone columns. Model development and validation involve using a database of 60 experimental samples from published studies. The input variables consisted of the length-to-diameter ratio (L/D), the area replacement ratio (As), number of stone columns (Ns) and undrained soil cohesion (Cu).The model performances were evaluated using statistical indices, as well as uncertainty, reliability, and sensitivity analyses. The results showed that SVR presented the best accuracy, with a correlation coefficient <i>R</i> of 0.941 and the lowest uncertainty (<i>U</i><sub><i>95</i></sub> = 8.377%) for the testing stage. Both the RF and GRNN also showed satisfactory predictive performance, whereas the RBNN model underperformed in comparison to the other models. Based on sensitivity analysis, the area replacement ratio (As) and undrained cohesion (Cu) were the most important parameters that affect bearing capacity. The results showed that ML methods, especially SVR, can offer accurate and efficient means of predicting the bearing capacity of floating stone columns and can assist in safer and resource-efficient, sustainable geotechnical designs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning Techniques for Accurate Prediction of Bearing Capacity of Foundation Resting on Soil Reinforced with Floating Stone Columns: A Comparative Performance Analysis

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Bashar H. Ismael,
  • Mohamed Khalid AlOmar,
  • Junied A. Bakr,
  • Mustafa Mohammed Aljumaily,
  • Adil Masood,
  • Siti Fatin Mohd Razali,
  • Saif Alzabeebee,
  • Osama Khamees Ali,
  • Mohammed Majeed Hameed

摘要

Stone column is an effective and sustainable technique of ground improvement, but the prediction of the bearing capacity of foundation built on stone column reinforced soils is still difficult to achieve because of uncertainties in the conventional method. In this study, the performance of four Machine Learning models namely Radial Basis Neural Network (RBNN), Generalized Regression Neural Network (GRNN), Random Forest (RF) and Support Vector Regression (SVR) is compared to develop a prediction model for the bearing capacity (qu) of floating stone columns. Model development and validation involve using a database of 60 experimental samples from published studies. The input variables consisted of the length-to-diameter ratio (L/D), the area replacement ratio (As), number of stone columns (Ns) and undrained soil cohesion (Cu).The model performances were evaluated using statistical indices, as well as uncertainty, reliability, and sensitivity analyses. The results showed that SVR presented the best accuracy, with a correlation coefficient R of 0.941 and the lowest uncertainty (U95 = 8.377%) for the testing stage. Both the RF and GRNN also showed satisfactory predictive performance, whereas the RBNN model underperformed in comparison to the other models. Based on sensitivity analysis, the area replacement ratio (As) and undrained cohesion (Cu) were the most important parameters that affect bearing capacity. The results showed that ML methods, especially SVR, can offer accurate and efficient means of predicting the bearing capacity of floating stone columns and can assist in safer and resource-efficient, sustainable geotechnical designs.